Search

Conselho Nacional de Desenvolvimento Cientifico e Tecnologico. Grant Number: 165009/2015-2 Fundacao de Amparo a Pesquisa do Estado de Sao Paulo (at CEMEAI). Grant Number: 2013/07375-0

There are many computational options nowadays, capable to serve properly a wide range of users. Web Services, Cloud Computing, Internet of Things among others are some of the ways technology can be offered. To scientific research this variety is fundamental. One of the major concerns in distributed environments such as cloud computing is the size of data and how to handle it's transport. In some cases, it is necessary to save information in hard drives and send it to other locations. The experiments in Protein Structure Predictions (PSP) usually results in very large amounts of data that are difficult to move from one place to another. Besides that, those results have different computational costs, processing time and price range. The current state-of-art shows the limitations the researchers have to face when working with local computers and most of their work is focused on improve the PSP algorithms. We understand that the historic runs in such systems are a reliable source to be studied and we propose a application of machine learning techniques that can be used as a base for a decision maker capable of defining machine configurations and a proper set-up, while keeping Quality of Service (QoS). The results of the kNN algorithm show a strategy capable of predicting with good accuracy the outcomes such as processing time and resulting file sizes. This defines the base for the decision maker mechanism.

Green computing has emerged as a hot topic leading to a need to understand energy consumption of computations. This need also extends to devices with limited resources as are common in the internet of things. RESTful services have shown their potential on such devices, but there are many choices of frameworks for their development and execution. Current research has analysed performance of the frameworks but no attention has been given to systematically studying their power consumption. In this paper we analyse the execution behaviour and power consumption of web services on devices with limited resources and make initial observations that should influence future development of web service frameworks. Specifically, we conduct experiments comparing web services in the Axis2 and CXF frameworks analysing the respective performance and power consumption. Bringing together the best features of small devices and SoC, it is possible to provide diverse, mobile and green applications — however careful selection of development environments can make significant differences in performance and energy consumption.

Fault tolerance techniques can improve the trust of users in service oriented architectures as they can ensure data availability. This paper presents an implementation of a novel fault tolerance mechanism in a SOA architecture which simultaneously provides increased availability and better quality of service. In addition to this mechanism, a service selector using reputation ratings of the architecture components is discussed. The selection is based on information from past transactions of the components of the architecture, which allows to identify the best web services able to meet the requests of customers. The mechanisms are tested and a performance evaluation is presented to validate the results.

This paper analyzes the execution behavior of web services on devices with limited resources. The experiments compare web services in the Axis2 and CXF frameworks analyzing performance and power consumption. To determine which framework is better suited for service provision, a testing environment and a performance and energy evaluation between them are presented. We show that the Raspberry Pi can be useful in service-oriented applications for different types of tasks. Bringing together the best features of small devices and SoC, it is possible to provide diverse, mobile and green applications.

Performing functionality testing in serviceoriented architectures is not a trivial task. The difficulty is especially the large number of components that may be present in a SOA such as brokers, providers, service registries, clients, monitoring tools, data storage tools, etc. Thus, in order to facilitate the process of conducting functional testing and capacity planning in service-oriented systems, we present PEESOS. This first version is a functional prototype that offers facilities to assist researchers and industry to test their new applications, allowing collaborations that can be done between the participants to achieve an appropriate objective when developing a new application. The first results show that it is possible to make a planning environment easier to operate and to readily obtain results for performance evaluation of a target architecture. Since this is a first version of the prototype, it has interface and scalability limitations as well as needing improvements in performance of the logs repository and also in a core engine. We hope that such limitations can be corrected in the near future, including gathering information from the scientific community to make the prototype a useful and accessible tool. PEESOS is on-line and available at http://peesos.wsarch.lasdpc.icmc.usp.br.